New research shows workers use AI but don’t trust it

We wanted to understand how people use generative AI at work and what challenges this brings. So we surveyed 289 Australian professionals from government, private-sector and not-for-profit organisations. We found respondents:

• use AI tools at work at all stages of the writing process and prefer Copilot
• can’t rely on AI outputs
• are rightly sceptical about AI
• need more training on writing with AI.

Want to know more about our key findings? Read on.

AI use at work is widespread

Our research found that almost 90% of respondents use generative AI at least once a month, with 1 in 3 respondents using it every day. Australian workplaces are aware of this: most respondents reported that their organisations already have a generative AI policy (57%). For organisations without a policy, 16% of respondents told us their organisations were developing one.

Workers use AI tools at all stages of the writing process and prefer Copilot

Respondents told us they rely on generative AI for a range of uses:

• 63% for refining their written expression
• 58% for summarising long texts
• 47% for brainstorming
• 33% for drafting emails and other documents
• 33% for researching.

Respondents overwhelmingly preferred Copilot (67%) and ChatGPT (35%), and were less likely to use Gemini, Grammarly and Claude. While 1 in 13 people said their workplace had its own in-house AI tool, we expect this to increase as more organisations invest in tailored AI tools.

Respondents can’t rely on AI outputs

Our research uncovered a range of problems with AI-generated output:

• 71% reported cliched language
• 59% reported hallucinations.

Others reported inappropriate tone, overly long prose and summaries that miss key information.

People are rightly sceptical about AI

We also asked how confident respondents felt about AI accuracy and reliability. Their answers ranged from very low (0% confidence) to very high (92%), with an average score of 45%. One respondent told us, ‘AI is still far too inaccurate and [unfortunately] real-world situations are being used as a testing ground.’

Looking to the future, we asked how optimistic workers felt about the potential impact of AI in the workplace. The average score was 53.5% – cautious optimism. One respondent commented that they were both excited and concerned. Excited because AI is the future and it has so much potential. Concerned because it could affect the skills of the future workforce. They also wondered how to ‘keep the nuance and humanity in the work’ and protect against bias.

Workers need more training on writing with AI

Despite the widespread uptake of generative AI, only 21.1% of respondents completed training on how to use it. Of those who hadn’t done any training yet, 2 in 5 wanted to do formal training and 1 in 3 were happy to experiment by themselves.

For formal training to be of any use, it needs to be relevant, focusing on the fundamentals of clear communication and helping people understand what good looks like. But most importantly, it needs to boost critical-thinking skills to ensure humans stay in control of the writing process and take responsibility for what they write.

And that is exactly what our plain language training does. Our writing tools come with critical thinking built in and slot perfectly into your workflow. Want better documents? Chat to us about our ISO-aligned writing tools.

Read Primary Comms Group’s media release about our research.